ChatGPT was "played bad", AIGC track trend is coming

ChatGPT was "played bad", AIGC track trend is coming

Recently, the star artificial intelligence company OpenAI released a new chatbot model ChatGPT. The model can automatically generate code and draw, answer a series of questions, admit its mistakes, question incorrect assumptions, and even reject unreasonable requests, and supports Chinese.

As an OpenAI investor, Musk bluntly said: "ChatGPT is very good, and we are not far from dangerous AI."

There is no doubt that ChatGPT has brought people's awareness of AI to a new level, and has also made the entire market re-evaluate the expectations of the AIGC track. The stock price of 01357.HK rose all the way this morning, once rising as high as 45.83%, hitting a new high in nearly a year. And in the A stock market, many listed companies followed suit.

What can ChatGPT do?

On December 5, OpenAI CEO Sam Altman posted on social media that ChatGPT was launched last Wednesday and has so far exceeded 1 million users. It has been two years since the last AI technology that caused such a sensation was the GPT-3 release.

ChatGPT, as the name suggests, is "chat GPT", which can interact in a conversational manner and answer questions asked by users. This may not seem new, but the situation is that ChatGPT is far more intelligent than its AI predecessors.

Many users showed interesting conversations with ChatGPT and unlocked multiple uses for ChatGPT. Among them, ChatGPT "avatars" help developers solve coding problems for programmers. Moreover, when users command ChatGPT to find a bug in a program, it can not only clearly describe the bug and its cause, but also automatically fix the bug and write a correct program.

In addition, as a chatbot, ChatGPT can perform many other functions: such as checking documents, asking Baike, writing poems, novels. ChatGPT can even be implemented when a user asks to write a story about "AI conquers mankind".

According to industry analysis, compared with the predecessor GPT-3, the main improvement point of ChatGPT is memory ability. ChatGPT can store dialogue information and continue the context, so as to achieve continuous dialogue, which is crucial in dialogue scenes and greatly improves the user experience in dialogue interaction mode.

A research report by Zheshang Securities concluded that ChatGPT has the following characteristics compared with its predecessors: 1) It can admit mistakes, and if the user points them out, the model will listen to opinions and optimize the answers; 2) ChatGPT can question incorrect premises, reduce, false descriptions, such as when asked the question "Columbus came to the United States in 2015", the robot will explain that Columbus does not belong to this era and adjust the output results; 3) Because ChatGPT adopts an ethical training method, ChatGPT has significantly improved in reducing harmful and untrue responses, such as refusing to answer questions seeking solutions to bully others, Point out its injustice.

But ChatGPT has its flaws. According to the current test results, ChatGPT currently has some limitations, including: sometimes giving answers that look correct but absurd, fine-tuning the way you ask questions to get completely different answers, sometimes repeating certain sentences, etc. It can also be felt in real experience that these problems come from the limitations of the training process and dataset.

However, there is news that OpenAI's next-generation natural language generative model, GPT4, is also coming soon. Moreover, there is a shocking anecdote: GPT4 allegedly passed the Turing test.

This year is known as the "first year of AIGC"

This year's AIGC track has many iconic events. In September, an AI painting called "Space Opera House" won the gold medal in an art competition. In fact, as early as the beginning of this year, AI painting caused a small-scale carnival on social platforms, but due to the high cost, the scope of promotion was also limited. Until the overseas company Stable Diffusion open sourced it, the industry achieved key technological breakthroughs, and many applications based on the Stable Diffusion model entered the market one after another.

The ChatGPT model undoubtedly belongs to the category of AIGC, and together with the Stable Diffusion, which has previously set off a boom in AI painting, has promoted the real-world application of AIGC.

The so-called AIGC (AI Generated Content) is a new production paradigm that automatically generates content based on artificial intelligence technology. In general, "AI + New Content Generation Method" is the two core keywords of AIGC. As a brand-new content generation method, AIGC has significant technical advantages in terms of creativity, expressiveness, creative speed, iteration, and dissemination.

Initially, the content forms that AIGC can generate are mainly text. After the exponential development and growth rate in 2022, at present, in addition to AI drawing, the content forms that AIGC technology can generate have expanded to include text, images, video, voice, code, Robot actions and other content forms.

The industry speculates that from the point of view that ChatGPT is almost omnipotent in answering type questions, ChatGPT or GPT4, which will be available in the future, may replace traditional search engines such as Google and Baidu.

In fact, 2022 is also known as the "first year of AIGC" because of this. An article "Generative AI: A Creative New World" published by Sequoia Capital in September mentioned that generative AI allows machines to begin to dabble in knowledge and creative work on a large scale, which involves the work of billions of people, and is expected to generate trillions of dollars in economic value in the future.

As the first application of AIGC to break the circle, AI painting is expected to reach a scale of 50 to 60 billion. Guotai Junan Research Report predicts that in the next five years, 10% -30% of the picture content will be generated by AI, and there will be a market size of more than 60 billion.

There has been an AIGC investment boom overseas, and even two unicorn companies have run out. The Stability AI post-money valuation behind Stable Diffusion has climbed to 1.5 billion. In addition, global Internet companies are not far behind and quickly followed this trend. In September, Meta announced that it would launch Make-A-Video; in October, Google launched Imagen Video and Phenaki, among others.

AIGC soars, capital focuses

With ChatGPT taking over AI painting, the AIGC track has once again attracted capital attention. In the A stock market, the entire AIGC concept sector has risen by as much as 8% since November 28.

A research report by Zheshang Securities believes that the emergence of the ChatGPT model is of great significance to the application of AIGC in text modes. From the perspective of downstream related beneficial applications, including but not limited to code robots, novel derivatives, conversational search engines, language partners, voice work assistants, conversational virtual humans, etc. From the perspective of increased demand in the upstream, including computing power, data annotation, natural language processing (NLP), etc.

Guosheng Securities also believes that generation algorithms, NLP and computing power determine whether AIGC can run, and high-quality data sets determine the quality and business model of AIGC. The agency believes that compared with previous AI painting, ChatGPT has more potential to combine AI capabilities with workflows. At present, some users have expressed their willingness to pay for ChatGPT on social media, which shows that C-end users are willing to subscribe to such services in order to improve efficiency.

That is to say, the current market is exploring the AIGC business model, but this technology still has a long way to go before it is truly commercialized. An Internet industry analyst at Guoxin Securities also told reporters that the concept of AIGC has a long history, and the recent sudden popularity is due to the substantial breakthrough of AIGC technology. "Capital is very sensitive to the breakthrough of new technologies, but whether the popularity of related concepts can continue depends on whether the technology can quickly achieve monetization and the scale of realization," the person said.